There's a folk proverb amongst engineers that “anyone can build a bridge that is strong, but only an engineer can build a bridge that is just strong enough.”
An engineer's perspective must incorporate a well-balanced design that just doesn’t look exotic or gets the job done but is also practical from the point-of-view of cost, manufacturability, durability, or other factors. Design optimization is the process that enables engineers to deliver the most efficient design.
Before moving ahead, let’s introduce some basic terminologies relevant to any optimization problem. Firstly, data regarding every possible design is required for the system, known as input parameters. Secondly, essential objectives of functionalities are required to be inserted into the system. Lastly, a set of constraint functions is required for the optimization process to determine the feasibility of any design. All these together determine the domain of optimization. To solve an optimization problem, the input data are tested for satisfying the design objectives and within constraint conditions. By employing optimization algorithms, one can find the optimal point(s) in the entire domain, which is represented as the maxima or minima value of the objective function.
These ideas can directly be extrapolated to design optimization. Unlike traditional design
processes, contemporary design processes include design optimization from the onset of the design cycle. This reduces the number of iterations and improves efficiency by integrating every factor in each step of the design lifecycle of a product or machine. For instance, design engineers can think about removing an unused feature of a product to save weight or find better material that can withstand real-world conditions and stresses to avoid failures. With digitalization emerging in the design process, it is expected that more and more real-world situations can be mimicked. This expectation makes engineering problems more complex as many factors need to be accounted for, which translates to multiple constraints and conflicting objective functions. This is where computing technology becomes relevant. To perform design optimization on computers is where computer-aided engineering (CAE) software comes into the picture. A specialized CAE software enables design optimization simulation using sophisticated algorithms to find the optimal designs. This design optimization simulation software is currently employed in many industries to ensure the development of products of high quality, reliability, and cost-effectiveness.
Image source - Gebisa, A. W., & Lemu, H. G. (2017, December)
Let's take the mobility sector to understand the applicability of design optimization simulation. The mobility sector consists of industries that require extensive engineering, viz., automotive, aerospace, maritime, and railway. A primary cost factor in the mobility sector is the increased weight of critical components. Companies want to deliver the most lightweight vehicle as it reduces fuel consumption and the cost of production. However, lightweight also means reduced durability of the product as parts can break or fail easily. So now, engineers face the challenge of designing a vehicle that must be strong enough to provide a sufficient factor of safety and reliability but lightweight enough to be economically feasible. This is where design optimization becomes essential. Due to the relevance of physics-based simulation in these cases, such as structural mechanics principles, specialized design optimization methods, such as Topology Optimization (TO), are needed.
The competitiveness of mobility and manufacturing sectors keeps pushing decision-makers to achieve higher efficiency in their products while also reducing the cost of operations and time-to-market. This results in a very challenging position for design engineering teams – they need to consider many constraints and objectives, making the design optimization problem harder while also delivering the results in a shorter time frame. It becomes a tug-of-war problem between decision-makers and engineering design teams. At the end of the day, time-to-market wins the argument, and the designers are left with no choice but to simplify their desired problem by leaving out many factors. This means highly efficient designs are out of the question, and in most cases, that’s okay until major recall issues appear.
You might ask - what’s the solution to this critical problem? I would say let’s first look at the root cause. The choice between performing high-accuracy simulations (in the case of design optimization, it means considering a wider design space by not making simplifications) and the turnaround time of simulation that needs to be made is because of a lack of enough computing resources. While stacking up more parallel processors in high-performance computing resources might be an intuitive solution, they soon will not scale up due to engineering and cost challenges. To overcome the bottleneck of computational resources, we can look at Quantum computers as they are the next-in-line computing technology. Quantum computers are new-age and powerful computers that work on very different computing principles. To leverage these computing principles, we require Quantum algorithms to solve our problems, such as design optimization. Luckily, optimization problems are something Quantum algorithms have an affinity for and are being explored in various sectors such as logistics, supply chain, etc. While the literature for engineering design optimization problems doesn’t exist, BosonQ Psi is changing that.
Our team has created Quantum-inspired optimization (QIO) algorithm. This algorithmic approach is a significant improvement on classical algorithms but can run on both classical and Quantum computers. Advantage? While we wait for Quantum computers' infrastructure to improve (circuit depth, qubit count, and noise), we can leverage the QIO algorithms to gain "industrial advantages" by running simulations on Quantum simulators that run on classical HPCs. The QIO algorithm has been integrated with our design optimization (DO) solver. The advantages we are seeing with internal test cases are exciting. Using a QIO-based DO solver, accounting for a large number of variables and constraints becomes feasible as the QI algorithm can swift through the entire design space in much less time and with a smaller number of computing resources compared to classical algorithms. This allows complex engineering problems to be solved without choosing between highly efficient designs and time-to-market. Hence, a win-win situation for the decision-makers and engineering designers.
About BosonQ Psi
BosonQ Psi is an IP-rich and fast-growing enterprise SaaS venture that builds premium simulation software powered by Quantum computing. Currently, it is building BQPhy, the world’s first Quantum-powered engineering simulation software for enterprise customers from the automotive, aerospace, and manufacturing industries. Its simulation capabilities utilize a hybrid infrastructure of quantum computers and classical high-performance computers (HPC) to highlight near-term value additions to our customers. Its quantum-based solution significantly reduces the time for high-accuracy simulation, which means products get faster into the market and reduce costs from production and recalls.